On Mon Jan 27 13:49:14 2025 +0000, eric pouech wrote:
the only use of keeping that string is to avoid to flush the cache when passing rescan + the same string as the previous one... I'm not convinced it's worth it
Removing this field and its associated checks would result in flushing the cache every time the scan is restarted, regardless of whether the mask has changed. This would represent something of a departure from the existing approach to caching, which is to reuse the cache wherever possible.
Given that flushing the cache involves making an RPC call to the Wineserver, there is the potential that a minor difference in behaviour here could have meaningful performance implications for applications that restart scans frequently (particularly if applications do so concurrently from multiple processes or threads, resulting in contention over the Wineserver RPC call). We'll conduct some performance analysis to determine the impact on relevant workloads and post the results here to inform further discussion.